D
Deleted member 537459
Yesterday I went in for my 3rd (so called booster) shot of anti-venom. I hope I'm good here as well... lest I speak.This should be good for another 34 pages. I'll get popcorn.
Had the second one, still caught covid for a second time shortly after that. Covid light apparently as it was only a little fever. When I was OK again I suddenly got severe diarrhoea and ****ed blood for 3 weeks. No doctor wanted to visit me. The hospital was also refusing me as it could be a risk to the elderly (new variant maybe).
Two months very ill but it went away. No shots for me anymore.
Two months very ill but it went away. No shots for me anymore.
I was looking for a "minimum priced" DAC as part of a system for some family members. The main criteria was that it would also satisfy my needs, not necessarily my wants. To that end the SMSL DO100 does a good job as a preamplifier with a DAC having USB, optical, coax and bluetooth inputs. The balanced XLR outputs can have advantages downstream with respect to noise/distortion. Notwithstanding the sonics the specifications are extraordinary. This means that the layouts, power supplies, analog sections, coupling components (if they. exist), etc., are of design and quality to support those spec.'s.There are cheap DACs that have better power supplies and analog sections than some expensive DACs. There are simple DIY DACs that outperform expensive DACs too. It may be hard to swallow but price and quality are not 1:1 related. One pays a large part of the sum for the brand name and the emotion that the object is good. All part of the insecurity of humans in general. Many need to hear from others what is good
Having built an extreme amount of DACs of all kinds I can tell that external DACs are a lottery. A cheap(er) DAC built into an audio player is way more effective than putting time and energy in a intrinsically very good external device connected with a weak interface. That is the bottleneck in a lot of cases. Less = more, also here.
I plan to connect the DAC to a Topping PA5. It has a gain control with two analog inputs hence it is possible to connect a turntable to the other inputs. This was also reviewed as highly regarded by ASR for spec reasons. Sonically it could easily fail against other devices designed by those with a greater knowledge of psycho acoustic phenomenon, I'm not sure yet. As far as aesthetics are concerned, it seems these devices are easy enough to hide, as the aesthetics is of more critical importance to my daughter at least.
Last edited:
D
Deleted member 537459
Sorry to disagree. Dither means adding noise. There isn't any reason to add noise to a 16 bit DAC unless you are reducing the bit depth starting from hi-res music, wich will result as data truncation (moreover someone prefers truncation instad of dithering).
TDA1541A is specified as -95dB THD+N (-90dB for the non-crowns versions).
Its linearity is 0.5 to 1 LSB.
Adding dither means reducing its linearity and increasing its noise, so that the THD will be buried in the noise.
A couple of bits of dithering mean a noise around -84dB (or less due to the DAC non-linearity), so that the -85dB THD of the I/V stage will be buried in the DAC noise.
D
Deleted member 537459
Practically, there is no evidence that 10 dB more second harmonic THD is more harmful than an added 10 dB of noise.
FFT does not clarify this, and the distinction between the two sources of error is often not easily identifiable.
In any case there isn't any valid reason to apply dither to a 16 bit DAC wich plays 16 bit music.
Maybe you all should read a little more about dither: https://www.masteringthemix.com/blo...much the,correlated" from the original signal. Its not simple and its use is essential for good audio when used correctly.
Am I to understand that you don't give a damn whether the error is noise, harmonic distortion or intermodulation products between multiples of the signal frequency and multiples of the sample rate, and that you therefore measure SINAD rather than harmonic distortion? If so, then I agree that there is a theoretical limit (like I wrote in the post that you quoted, https://www.diyaudio.com/community/...quality-between-some-dacs.386815/post-7045860 ).
Indeed. The dither should be applied when the 16 bit music or 16 bit test signal is recorded, just before quantization to 16 bit. There is no point adding it after quantization.
Last edited:
I am out sick for a few days, but this looks like a train wreck starting to unfold.
Marcel, it looks to me like The Well Audio is thinking in terms of NOS dacs, where the problems with dither applied in a dac are the same as described by the mastering engineer article. I believe The Well Audio was just trying to explain why its a problem (and, please remember there is still a language barrier in play as further explained below).
(Aside: In the case of mastering 'dither' it isn't even necessary that the dithering agent be noise, nor was it noise in the case of UV-22 dither.)
The other thing I would like to mention is that the terms 'buried in noise,' and 'replaced by noise' both have their place in the process of using mastering dither and then truncation. The bits below bit-15 do get buried in noise since that is the only way to preserve them at all above the truncation point. Again, language may play a role here in choice and precise meaning of of English language terminology.
Regarding your own point of view, your understanding of dither in OS dac modulator design or in general is not being questioned. It seems like the confusion here in the recent conversion surrounds a question of context, dacs or mastering? Which one is it assumed everyone is talking about? Why assume talk is about mastering in a thread comparing the sound of various dac architectures?
Okay, thank you for your understanding. I'm still sick, and out of here again.
Marcel, it looks to me like The Well Audio is thinking in terms of NOS dacs, where the problems with dither applied in a dac are the same as described by the mastering engineer article. I believe The Well Audio was just trying to explain why its a problem (and, please remember there is still a language barrier in play as further explained below).
(Aside: In the case of mastering 'dither' it isn't even necessary that the dithering agent be noise, nor was it noise in the case of UV-22 dither.)
The other thing I would like to mention is that the terms 'buried in noise,' and 'replaced by noise' both have their place in the process of using mastering dither and then truncation. The bits below bit-15 do get buried in noise since that is the only way to preserve them at all above the truncation point. Again, language may play a role here in choice and precise meaning of of English language terminology.
Regarding your own point of view, your understanding of dither in OS dac modulator design or in general is not being questioned. It seems like the confusion here in the recent conversion surrounds a question of context, dacs or mastering? Which one is it assumed everyone is talking about? Why assume talk is about mastering in a thread comparing the sound of various dac architectures?
Okay, thank you for your understanding. I'm still sick, and out of here again.
D
Deleted member 537459
Thanks for clarifying, indeed the topic was reproduction and not mastering.
In order to better clarify, The Well Audio designers think that dither can also be applied after quantization during playback to avoid truncation of data, for example when playing 24 bit material with a 16 bit DAC. Indeed the TWSAFB-LT FIFO buffer provides dithering options, allowing real time comparison of the reproduction without and with a few kind of dither. There is also an option called "Adaptive dither" wich automatically disables the selected dithering option when not needed (no data truncation).
Well, this will appeal audiophilitics: "The ears are the most precise instrument to evaluate sound", "there are things we can hear but cannot measure", "hand built and discrete is always better" and similar nonsense.
Regarding harmonic distortion:
By definition, harmonic distortion is the ratio of the RMS value of the harmonics divided by the RMS value of the fundamental, converted to a percentage or expressed in dB if you like. So you apply a sine and measure at the frequency of the sine and its multiples.
As a thought experiment, suppose you would have a perfect 16 bit DAC with perfect reconstruction filter and wanted to measure its harmonic distortion at 1 kHz signal frequency and 44.1 kHz sample rate.
You would need a digital test signal. If you were to generate the test signal by just calculating sine values and rounding them to 16 bits, you would get distortion products due to the staircase-shaped response of the rounding (quantizing) function. The 1st...22nd harmonic would be below the Nyquist frequency, higher harmonics would alias to multiples of 100 Hz (this being the beat frequency of 1 kHz and 44.1 kHz; everything repeats every 10 ms (1)).
By definition, you would measure in narrow bands around 1 kHz, 2 kHz, 3 kHz ... 22 kHz, find there is nothing at 23 kHz and higher due to the perfect reconstruction filter (and even if there were, it is usual to ignore anything above Nyquist anyway) and you would ignore the spectral peaks at multiples of 100 Hz that are not multiples of 1 kHz. You would end up with some nonzero distortion figure, but it is really the distortion of the rounding to 16 bits that you measure.
If you generate the test signal by calculating sine samples, adding dither and then rounding to 16 bits, making sure that the sum of the sine and the dither doesn't cause clipping, then the quantization distortion would turn into a kind of noise. You can then get the measured harmonic distortion as low as you like by measuring in very narrow bands around 1 kHz, 2 kHz ... 22 kHz - as long as the DAC is perfect, anyway.
On the other hand, for a SINAD measurement, you would just measure at 1 kHz and over the whole band except 1 kHz (whole band meaning 0 to Nyquist or the audio band, with or without A-weighting) and divide the RMS level at 1 kHz by the RMS level over the whole band except 1 kHz. You then get a finite value with and without dither.
(1): This also means that you don't test all 65536 DAC codes. You can solve that by choosing a slightly different test frequency.
By definition, harmonic distortion is the ratio of the RMS value of the harmonics divided by the RMS value of the fundamental, converted to a percentage or expressed in dB if you like. So you apply a sine and measure at the frequency of the sine and its multiples.
As a thought experiment, suppose you would have a perfect 16 bit DAC with perfect reconstruction filter and wanted to measure its harmonic distortion at 1 kHz signal frequency and 44.1 kHz sample rate.
You would need a digital test signal. If you were to generate the test signal by just calculating sine values and rounding them to 16 bits, you would get distortion products due to the staircase-shaped response of the rounding (quantizing) function. The 1st...22nd harmonic would be below the Nyquist frequency, higher harmonics would alias to multiples of 100 Hz (this being the beat frequency of 1 kHz and 44.1 kHz; everything repeats every 10 ms (1)).
By definition, you would measure in narrow bands around 1 kHz, 2 kHz, 3 kHz ... 22 kHz, find there is nothing at 23 kHz and higher due to the perfect reconstruction filter (and even if there were, it is usual to ignore anything above Nyquist anyway) and you would ignore the spectral peaks at multiples of 100 Hz that are not multiples of 1 kHz. You would end up with some nonzero distortion figure, but it is really the distortion of the rounding to 16 bits that you measure.
If you generate the test signal by calculating sine samples, adding dither and then rounding to 16 bits, making sure that the sum of the sine and the dither doesn't cause clipping, then the quantization distortion would turn into a kind of noise. You can then get the measured harmonic distortion as low as you like by measuring in very narrow bands around 1 kHz, 2 kHz ... 22 kHz - as long as the DAC is perfect, anyway.
On the other hand, for a SINAD measurement, you would just measure at 1 kHz and over the whole band except 1 kHz (whole band meaning 0 to Nyquist or the audio band, with or without A-weighting) and divide the RMS level at 1 kHz by the RMS level over the whole band except 1 kHz. You then get a finite value with and without dither.
(1): This also means that you don't test all 65536 DAC codes. You can solve that by choosing a slightly different test frequency.
- Status
- Not open for further replies.